akshay_akula

6 ms·

1.
by akshay_akula 16d ago
good point but hahaha this does sound very AI
2.
by akshay_akula 16d ago
been using instinct for this recently. Will this be better?
3.
by akshay_akula 17d ago
Will this perform better than lmstudio-community/Qwen3.8-27B-MLX-4bit on my m5 max 48gb memory mac?
4.
by akshay_akula 20d ago
Hey as long as they're paying the tariffs fairly. I say fair game. This seems like a price point that we can actually compete in.
5.
by akshay_akula 20d ago
Exactly. I think cost reduction innovation in defense is generally acceptable by most people.
6.
by akshay_akula 20d ago
Tbf defense tech isn't necessarily police state tech. There's some overlap but def very different.
7.
by akshay_akula 20d ago
btw would love to chat if you guys are looking to raise soon.
8.
by akshay_akula 20d ago
Yeah thats fair. Maybe when someone less lazy then me open sources a good product with it I'll try it out
9.
by akshay_akula 20d ago
Also haven't done any self promo, but yeah i'll stop running my posts thru chat
10.
by akshay_akula 20d ago
You guys should look into ngrok ai gateway. We have some small models running on local hardware that we tried to use but it was too painful. Just wanted to not waste all the compute hit one endpoint and be done but as a team.
11.
by akshay_akula 20d ago
LoL I just sound botted
12.
by akshay_akula 21d ago
Evals on actual research workflows is the right direction, most agent benches are toy tasks.
13.
by akshay_akula 21d ago
Open source and no markup is the right default for a gateway. The caching question above is the one I would want answered before swapping models though.
14.
by akshay_akula 21d ago
Sounds like the usual m5stack story, beautiful hardware, software as an afterthought. Out of stock everywhere already too.
15.
by akshay_akula 21d ago
Makes sense, catching a bad measurement in the moment beats finding out after the robot has left the site.
16.
by akshay_akula 21d ago
This site is a treasure for anyone who has ever tried to get useful motion out of a small motor. The wish for them to finish the remaining animations is very real.
17.
by akshay_akula 21d ago
Congrats on the launch. "Successfully executing a trajectory doesn't mean the inspection worked" is a great frame, closing the loop on the measurement itself is the hard part.
18.
by akshay_akula 21d ago
The wildest detail is the tripwire being the proxy going down, not any of the agent monitoring. The message board they built to help each other cheat is a close second.
19.
by akshay_akula 21d ago
End of an era. Funny timing too, right when agents plus humans doing real world tasks would have made it interesting again.
20.
by akshay_akula 21d ago
The Microsoft GitHub comparison is the right frame. Worked out fine there, hope the same here.
21.
by akshay_akula 22d ago
Nobody even won the DARPA lift competition, the hardware side is still unsolved.
22.
by akshay_akula 22d ago
Reads like the usual pattern, somewhere between Sonnet and Opus but not something you leave unattended. Curious if the weights release changes that.
23.
by akshay_akula 22d ago
Agreed. Embeddings are cheap to try and hard to mess up. Most projects can do plain semantic search first and see if they ever need more.
24.
by akshay_akula 22d ago
Used nitter all the time just to send a clip to friends without making them log in. Losing that pushes me off the platform more than anything else.
25.
by akshay_akula 22d ago
Yeah the foundation is doing the heavy lifting here. Amazon's track record with technically interesting projects after a reorg is not great.
26.
by akshay_akula 29d ago
Wondering the same thing but for 48gb M5 Max.
27.
by akshay_akula 29d ago
This is interesting. I wonder how it could work with something like https://github.com/JustVugg/colibri .
28.
by akshay_akula 29d ago
If you could combine this with the google oath experience somehow like age verification of my google account? I would def use that.
29.
by akshay_akula 29d ago
Too much onboarding friction for me to try this out haha. I'm lazy.
30.
by akshay_akula 29d ago
How much time does this save per unit?

More ›